Kasra Rouhi
@kasrarouhi.bsky.social
Ph.D., Electrical Engineering and Computer Science
University of California, Irvine
I am interested in #physics, #astronomy, #computer_science, and #psychology
University of California, Irvine
I am interested in #physics, #astronomy, #computer_science, and #psychology
2025 saw 19% internet traffic growth, mobile hit 43%, AI bots made up 4.2% of HTML requests, and post-quantum TLS hit 52%. The web is evolving fast and AI is a huge part of it. Expect even more shifts in how we connect, and scale online. 📊🌐📱🤖
#InternetTrends2025
radar.cloudflare.com/year-in-revi...
#InternetTrends2025
radar.cloudflare.com/year-in-revi...
December 17, 2025 at 6:22 AM
2025 saw 19% internet traffic growth, mobile hit 43%, AI bots made up 4.2% of HTML requests, and post-quantum TLS hit 52%. The web is evolving fast and AI is a huge part of it. Expect even more shifts in how we connect, and scale online. 📊🌐📱🤖
#InternetTrends2025
radar.cloudflare.com/year-in-revi...
#InternetTrends2025
radar.cloudflare.com/year-in-revi...
A recent study found that at least 13.5% of biomedical papers in 2024 were written with the help of LLMs like ChatGPT, a shift even larger than COVID's impact on scientific language. The AI era in academia is here and accelerating. 🚀📚🧬📊
#AIAcademia #ChatGPT #Biomed
www.science.org/doi/10.1126/...
#AIAcademia #ChatGPT #Biomed
www.science.org/doi/10.1126/...
December 4, 2025 at 9:00 AM
A recent study found that at least 13.5% of biomedical papers in 2024 were written with the help of LLMs like ChatGPT, a shift even larger than COVID's impact on scientific language. The AI era in academia is here and accelerating. 🚀📚🧬📊
#AIAcademia #ChatGPT #Biomed
www.science.org/doi/10.1126/...
#AIAcademia #ChatGPT #Biomed
www.science.org/doi/10.1126/...
Relying on AI like ChatGPT for learning may lead to shallower knowledge, says a new study. Active engagement—like digging through Google links—builds deeper understanding. Even when the facts are the same, passive reading weakens retention. 🧠✨📚
#AI #EdTech #ChatGPT
academic.oup.com/pnasnexus/ar...
#AI #EdTech #ChatGPT
academic.oup.com/pnasnexus/ar...
December 4, 2025 at 12:30 AM
Relying on AI like ChatGPT for learning may lead to shallower knowledge, says a new study. Active engagement—like digging through Google links—builds deeper understanding. Even when the facts are the same, passive reading weakens retention. 🧠✨📚
#AI #EdTech #ChatGPT
academic.oup.com/pnasnexus/ar...
#AI #EdTech #ChatGPT
academic.oup.com/pnasnexus/ar...
AI helps simulate the Milky Way in stunning detail—100 billion stars are mapped at 100× speed! By fusing deep learning with HPC, researchers shrank 36 years of computing into 115 days. A revolutionary era in galaxy modeling is here. 🚀🌌💥
#AIAstronomy #MilkyWay
dl.acm.org/doi/10.1145/...
#AIAstronomy #MilkyWay
dl.acm.org/doi/10.1145/...
November 27, 2025 at 8:20 AM
AI helps simulate the Milky Way in stunning detail—100 billion stars are mapped at 100× speed! By fusing deep learning with HPC, researchers shrank 36 years of computing into 115 days. A revolutionary era in galaxy modeling is here. 🚀🌌💥
#AIAstronomy #MilkyWay
dl.acm.org/doi/10.1145/...
#AIAstronomy #MilkyWay
dl.acm.org/doi/10.1145/...
Major leap for the quantum internet: UChicago scientists extended quantum link distance from kilometers to over 2,000 km using ultra-pure crystals built atom by atom. Quantum computers could soon connect across continents, paving the way for a quantum network. 🌍⚛️
#QuantumTech
doi.org/10.1038/s414...
#QuantumTech
doi.org/10.1038/s414...
Dual epitaxial telecom spin-photon interfaces with long-lived coherence - Nature Communications
Erbium quantum emitters operating in the telecom C band are promising for spin-photon interfaces but achieving both optical and spin coherence has been challenging. Gupta et al. report two types of er...
www.nature.com
November 26, 2025 at 7:55 PM
Major leap for the quantum internet: UChicago scientists extended quantum link distance from kilometers to over 2,000 km using ultra-pure crystals built atom by atom. Quantum computers could soon connect across continents, paving the way for a quantum network. 🌍⚛️
#QuantumTech
doi.org/10.1038/s414...
#QuantumTech
doi.org/10.1038/s414...
New research shows LLMs can fuse audio & motion data to recognize activities like sports & household tasks—zero-shot! No extra training or memory needed. A smart step for multimodal AI. 🤖📱💡
#AIResearch #ActivityRecognition
www.arxiv.org/abs/2509.10729
#AIResearch #ActivityRecognition
www.arxiv.org/abs/2509.10729
November 26, 2025 at 7:50 PM
New research shows LLMs can fuse audio & motion data to recognize activities like sports & household tasks—zero-shot! No extra training or memory needed. A smart step for multimodal AI. 🤖📱💡
#AIResearch #ActivityRecognition
www.arxiv.org/abs/2509.10729
#AIResearch #ActivityRecognition
www.arxiv.org/abs/2509.10729
AI talent is scarce, but Europe is rethinking education to fix it. From teacher training with AI in Manchester to entrepreneurship programs like AI-ENTR4YOUTH, a new wave of AI-native learners is rising. 🚀📉 🧪
#FutureOfWork #AITalent
www.artificialintelligence-news.com/news/ai-educ...
#FutureOfWork #AITalent
www.artificialintelligence-news.com/news/ai-educ...
What Europe's AI education experiments can teach a business
In Europe, the use of AI in education is being led by several experimental programmes that can teach business leaders sound lessons.
www.artificialintelligence-news.com
November 21, 2025 at 2:24 AM
AI talent is scarce, but Europe is rethinking education to fix it. From teacher training with AI in Manchester to entrepreneurship programs like AI-ENTR4YOUTH, a new wave of AI-native learners is rising. 🚀📉 🧪
#FutureOfWork #AITalent
www.artificialintelligence-news.com/news/ai-educ...
#FutureOfWork #AITalent
www.artificialintelligence-news.com/news/ai-educ...
Google’s new Quantum Echoes algorithm on the Willow chip achieves verifiable quantum advantage — running 13,000x faster than top classical supercomputers. A big leap toward real-world quantum apps in chemistry, medicine & materials. 🧪💡⚛️
#QuantumComputing
blog.google/technology/r...
#QuantumComputing
blog.google/technology/r...
November 5, 2025 at 12:28 AM
Google’s new Quantum Echoes algorithm on the Willow chip achieves verifiable quantum advantage — running 13,000x faster than top classical supercomputers. A big leap toward real-world quantum apps in chemistry, medicine & materials. 🧪💡⚛️
#QuantumComputing
blog.google/technology/r...
#QuantumComputing
blog.google/technology/r...
AI is reshaping business messaging — and also making it harder to govern. LeapXpert brings structure, oversight, and Communication Data Intelligence to AI-driven chat across WhatsApp, Teams, iMessage & more. Smarter messages, safer systems. 🔐🤖📱
#AI
www.artificialintelligence-news.com/news/how-lea...
#AI
www.artificialintelligence-news.com/news/how-lea...
November 2, 2025 at 6:35 AM
AI is reshaping business messaging — and also making it harder to govern. LeapXpert brings structure, oversight, and Communication Data Intelligence to AI-driven chat across WhatsApp, Teams, iMessage & more. Smarter messages, safer systems. 🔐🤖📱
#AI
www.artificialintelligence-news.com/news/how-lea...
#AI
www.artificialintelligence-news.com/news/how-lea...
New study proposes the LLM Brain Rot Hypothesis: training on junk web text (like low-quality Twitter posts) causes lasting cognitive decline in language models — hurting reasoning, safety, and boosting dark traits. Quality > quantity in AI training! 🚫🧵🧠
#AIHealthCheck
www.arxiv.org/abs/2510.13928
#AIHealthCheck
www.arxiv.org/abs/2510.13928
October 24, 2025 at 5:12 AM
New study proposes the LLM Brain Rot Hypothesis: training on junk web text (like low-quality Twitter posts) causes lasting cognitive decline in language models — hurting reasoning, safety, and boosting dark traits. Quality > quantity in AI training! 🚫🧵🧠
#AIHealthCheck
www.arxiv.org/abs/2510.13928
#AIHealthCheck
www.arxiv.org/abs/2510.13928
Apple unveils FS-DFM — a diffusion-based language model that writes long text up to 128x faster than rivals! It generates high-quality text in just 8 steps vs 1000+, with lower perplexity and stable entropy. A major leap for efficient LLMs! 🔍📉🚀🧠💬
#AIAcceleration
arxiv.org/abs/2509.20624
#AIAcceleration
arxiv.org/abs/2509.20624
October 17, 2025 at 5:56 AM
Apple unveils FS-DFM — a diffusion-based language model that writes long text up to 128x faster than rivals! It generates high-quality text in just 8 steps vs 1000+, with lower perplexity and stable entropy. A major leap for efficient LLMs! 🔍📉🚀🧠💬
#AIAcceleration
arxiv.org/abs/2509.20624
#AIAcceleration
arxiv.org/abs/2509.20624
Apple's M5 chip delivers 4x AI GPU performance over M4, packs Neural Accelerators in each core, a faster Neural Engine, the world’s fastest CPU core, and 30% more memory bandwidth. Available now for MacBook Pro, iPad Pro, & Vision Pro. 🍏⚙️💡
#AppleSilicon #M5 #AIChip
www.apple.com/newsroom/202...
#AppleSilicon #M5 #AIChip
www.apple.com/newsroom/202...
October 15, 2025 at 8:44 PM
Apple's M5 chip delivers 4x AI GPU performance over M4, packs Neural Accelerators in each core, a faster Neural Engine, the world’s fastest CPU core, and 30% more memory bandwidth. Available now for MacBook Pro, iPad Pro, & Vision Pro. 🍏⚙️💡
#AppleSilicon #M5 #AIChip
www.apple.com/newsroom/202...
#AppleSilicon #M5 #AIChip
www.apple.com/newsroom/202...
LLMs don’t lie on purpose—they guess when unsure because today’s benchmarks reward confident answers, not admitting uncertainty. Hallucinations arise from this misalignment. To fix AI trustworthiness, we must rethink how we train and evaluate these models. 🔍 🧠🤖
arxiv.org/abs/2509.04664
arxiv.org/abs/2509.04664
October 11, 2025 at 9:42 PM
LLMs don’t lie on purpose—they guess when unsure because today’s benchmarks reward confident answers, not admitting uncertainty. Hallucinations arise from this misalignment. To fix AI trustworthiness, we must rethink how we train and evaluate these models. 🔍 🧠🤖
arxiv.org/abs/2509.04664
arxiv.org/abs/2509.04664
Harvard & MIT physicists unveiled a quantum leap: a 3,000+ qubit system that can run continuously for over 2 hours—solving a major bottleneck in quantum computing. For real-world quantum power, lasers replace lost atoms on the fly. 🧠💡⚛️🚀🔬
#QuantumComputing #HarvardMIT
www.nature.com/articles/s41...
#QuantumComputing #HarvardMIT
www.nature.com/articles/s41...
October 11, 2025 at 8:18 AM
Harvard & MIT physicists unveiled a quantum leap: a 3,000+ qubit system that can run continuously for over 2 hours—solving a major bottleneck in quantum computing. For real-world quantum power, lasers replace lost atoms on the fly. 🧠💡⚛️🚀🔬
#QuantumComputing #HarvardMIT
www.nature.com/articles/s41...
#QuantumComputing #HarvardMIT
www.nature.com/articles/s41...
A minimalist slice of matcha-citrus cake on a golden fork, suspended in elegance against a muted green—clean, calm, and irresistibly refined. 🍰💚✨
#LeonardoAI
#LeonardoAI
October 11, 2025 at 4:08 AM
A minimalist slice of matcha-citrus cake on a golden fork, suspended in elegance against a muted green—clean, calm, and irresistibly refined. 🍰💚✨
#LeonardoAI
#LeonardoAI
Red Hat’s latest survey reveals a striking truth: 89% of UK businesses haven’t seen customer value from AI yet, but AI investment is set to grow by 32% by 2026. Priorities? AI + security top the list. 🧠🔒💷
#AIinUK #OpenSourceAI
www.artificialintelligence-news.com/news/ai-valu...
#AIinUK #OpenSourceAI
www.artificialintelligence-news.com/news/ai-valu...
October 11, 2025 at 4:04 AM
Red Hat’s latest survey reveals a striking truth: 89% of UK businesses haven’t seen customer value from AI yet, but AI investment is set to grow by 32% by 2026. Priorities? AI + security top the list. 🧠🔒💷
#AIinUK #OpenSourceAI
www.artificialintelligence-news.com/news/ai-valu...
#AIinUK #OpenSourceAI
www.artificialintelligence-news.com/news/ai-valu...
World’s most accurate artificial tongue is here — using graphene + AI, scientists have created a device that can detect flavors with 98.5% accuracy, even in moist environments! Big leap for food safety, healthcare, and robotics. 👅 🧪🤖
#AITaste
#Graphene
#BioInspiredTech
www.pnas.org/doi/10.1073/...
#AITaste
#Graphene
#BioInspiredTech
www.pnas.org/doi/10.1073/...
September 16, 2025 at 12:31 AM
World’s most accurate artificial tongue is here — using graphene + AI, scientists have created a device that can detect flavors with 98.5% accuracy, even in moist environments! Big leap for food safety, healthcare, and robotics. 👅 🧪🤖
#AITaste
#Graphene
#BioInspiredTech
www.pnas.org/doi/10.1073/...
#AITaste
#Graphene
#BioInspiredTech
www.pnas.org/doi/10.1073/...
Neuromorphic leap for drug discovery! SpiNNcloud sold a 4,320‑chip brain‑inspired supercomputer to University of Leipzig to speed protein folding for personalized medicine—faster & greener than GPUs. ⚡🧬🧠💊
#NeuromorphicAI
#DrugDiscovery
#SpiNNcloud
www.eetimes.com/spinncloud-s...
#NeuromorphicAI
#DrugDiscovery
#SpiNNcloud
www.eetimes.com/spinncloud-s...
August 5, 2025 at 11:39 PM
Neuromorphic leap for drug discovery! SpiNNcloud sold a 4,320‑chip brain‑inspired supercomputer to University of Leipzig to speed protein folding for personalized medicine—faster & greener than GPUs. ⚡🧬🧠💊
#NeuromorphicAI
#DrugDiscovery
#SpiNNcloud
www.eetimes.com/spinncloud-s...
#NeuromorphicAI
#DrugDiscovery
#SpiNNcloud
www.eetimes.com/spinncloud-s...
As AI gets smarter, it hallucinates more—fabricating facts with confidence. Is it a bug… or a feature? Experts say creative AI needs some imagination, but unchecked errors pose serious risks. Time to rethink trust & add guardrails. 🧠⚠️🔍
#LLMethics
#TruthInAI
www.livescience.com/technology/a...
#LLMethics
#TruthInAI
www.livescience.com/technology/a...
June 28, 2025 at 1:23 AM
As AI gets smarter, it hallucinates more—fabricating facts with confidence. Is it a bug… or a feature? Experts say creative AI needs some imagination, but unchecked errors pose serious risks. Time to rethink trust & add guardrails. 🧠⚠️🔍
#LLMethics
#TruthInAI
www.livescience.com/technology/a...
#LLMethics
#TruthInAI
www.livescience.com/technology/a...
Miniature orange AI kittens in kimonos, dead serious, mastering sushi with chopsticks in a serene Japanese room—8K cuteness meets culinary focus. 🍣🐾🎌
#LeonardoAI
#LeonardoAI
June 28, 2025 at 1:17 AM
Miniature orange AI kittens in kimonos, dead serious, mastering sushi with chopsticks in a serene Japanese room—8K cuteness meets culinary focus. 🍣🐾🎌
#LeonardoAI
#LeonardoAI
Are chatbots just telling us what we want to hear? Daniel Tunkelang explores the tension between usefulness and truth in AI. If we want honest answers, we must reward discomfort—not flattery. 🧠⚖️🤖💬
#AIethics
#ConfirmationBias
#ChatGPT
www.wsj.com/tech/ai/ai-c...
#AIethics
#ConfirmationBias
#ChatGPT
www.wsj.com/tech/ai/ai-c...
That Chatbot May Just Be Telling You What You Want to Hear
Artificial-intelligence chatbots tend to flatter users and be overly agreeable. It may feel good, but you may pay a price for such high praise.
www.wsj.com
June 25, 2025 at 8:38 PM
Are chatbots just telling us what we want to hear? Daniel Tunkelang explores the tension between usefulness and truth in AI. If we want honest answers, we must reward discomfort—not flattery. 🧠⚖️🤖💬
#AIethics
#ConfirmationBias
#ChatGPT
www.wsj.com/tech/ai/ai-c...
#AIethics
#ConfirmationBias
#ChatGPT
www.wsj.com/tech/ai/ai-c...
Meet the tiny Linux computer that's smaller than a passport photo! At just 40×35mm, it runs Ubuntu, has a screen, motion sensor, mic, Wi-Fi, BT, and even a GPU—all in your palm. The future of mini computing is adorable and powerful! 📦💡🖥️
#Linux
#TinyTech
#DevBoardGoals
www.youtube.com/watch?v=RhVB...
#Linux
#TinyTech
#DevBoardGoals
www.youtube.com/watch?v=RhVB...
Tiny Linux Computer
YouTube video by Coding Scientist
www.youtube.com
June 14, 2025 at 7:48 AM
Meet the tiny Linux computer that's smaller than a passport photo! At just 40×35mm, it runs Ubuntu, has a screen, motion sensor, mic, Wi-Fi, BT, and even a GPU—all in your palm. The future of mini computing is adorable and powerful! 📦💡🖥️
#Linux
#TinyTech
#DevBoardGoals
www.youtube.com/watch?v=RhVB...
#Linux
#TinyTech
#DevBoardGoals
www.youtube.com/watch?v=RhVB...
New world record! Japan hits 1 million GB/sec internet speed over 1,100 miles using a 19-core optical fiber—thin as today’s cables. Enough to download 10,000 4K movies in a second. The future of data is already here. 🧠⚡📡🚀
#FiberFuture
#1Petabit
#InternetSpeed
opg.optica.org/abstract.cfm...
#FiberFuture
#1Petabit
#InternetSpeed
opg.optica.org/abstract.cfm...
June 8, 2025 at 10:09 PM
New world record! Japan hits 1 million GB/sec internet speed over 1,100 miles using a 19-core optical fiber—thin as today’s cables. Enough to download 10,000 4K movies in a second. The future of data is already here. 🧠⚡📡🚀
#FiberFuture
#1Petabit
#InternetSpeed
opg.optica.org/abstract.cfm...
#FiberFuture
#1Petabit
#InternetSpeed
opg.optica.org/abstract.cfm...
Super-Turing AI mimics the brain to slash energy use! Researchers at Texas A&M have created a smarter, greener AI by combining memory + learning like neurons do—potentially revolutionizing how AI is built. 🌱🤖💡🧠⚡️
#SustainableAI
#SuperTuring
#BrainInspiredAI
www.science.org/doi/10.1126/...
#SustainableAI
#SuperTuring
#BrainInspiredAI
www.science.org/doi/10.1126/...
HfZrO-based synaptic resistor circuit for a Super-Turing intelligent system
Synaptic resistor circuits for Super-Turing AI systems feature low power consumption, fast learning, and agile adaptability.
www.science.org
May 18, 2025 at 2:20 AM
Super-Turing AI mimics the brain to slash energy use! Researchers at Texas A&M have created a smarter, greener AI by combining memory + learning like neurons do—potentially revolutionizing how AI is built. 🌱🤖💡🧠⚡️
#SustainableAI
#SuperTuring
#BrainInspiredAI
www.science.org/doi/10.1126/...
#SustainableAI
#SuperTuring
#BrainInspiredAI
www.science.org/doi/10.1126/...
Humanity may hit the singularity within the next 12 months, say some AI experts. A new study shows predictions shifting dramatically post-LLM boom, with AGI now expected as soon as 2030—or even sooner.
#Singularity
#AGI
#AIRevolution 🤖⌛🧠
www.popularmechanics.com/science/a639...
#Singularity
#AGI
#AIRevolution 🤖⌛🧠
www.popularmechanics.com/science/a639...
Humanity May Achieve the Singularity Within the Next 12 Months, Scientists Suggest
Predictions across the field range from a few months to a few decades, but experts agree—change is coming.
www.popularmechanics.com
May 18, 2025 at 2:07 AM
Humanity may hit the singularity within the next 12 months, say some AI experts. A new study shows predictions shifting dramatically post-LLM boom, with AGI now expected as soon as 2030—or even sooner.
#Singularity
#AGI
#AIRevolution 🤖⌛🧠
www.popularmechanics.com/science/a639...
#Singularity
#AGI
#AIRevolution 🤖⌛🧠
www.popularmechanics.com/science/a639...